Reviews
How to Read Degree Outcomes Before Choosing a Program
Degree outcomes show graduates' results, not what a program caused. Learn to assess employment and salary claims, read data, and compare your alternative.
Every program that publishes degree outcomes data is inviting you to make a causal inference: graduates of this program do well, therefore this program will make you do well. That inference is the hardest one in the whole field, almost nobody has the evidence to support it, and the numbers on the page are usually not even trying to.
This page names no program and quotes no published outcome figure of its own. It is about how to read the numbers you will be shown, and how to price the alternative you already have.
What to take away
- Government and public data sources are worth using, but they carry constraints you have to respect.
- The instinct is to compare a program's numbers against a national average, or against another program's numbers.
- The comparison that means something is against what happens if you do not do this, and you can price that yourself.
The causal problem, stated plainly
Programs select their students. That single fact does most of the damage.
Admitted students to a competitive program were, on average, already more likely to land good jobs than rejected ones, because they had the credentials, experience, network, persistence, and often the resources that won admission.
So when a program's graduates do well, at least three things are contributing: what the program taught, what the program's hiring channel provided, and who the program admitted. The published number does not separate them, and typically the third is doing a great deal of the work.
A program cannot report what its graduates would have earned had they not attended, because that group does not exist.
To establish the causal claim properly, compare admitted students with near-identical students who were not admitted, then follow them over years. Such a study is rare, expensive, context-specific, and not what a marketing page is built on.
What follows for you: treat outcome data as a description of a population, never as a forecast for a person.
What the definitions conceal
Before an outcome figure means anything, four questions have to be answered, and most reports answer none of them on the page where the figure appears.
Four questions before any figure
- Who counted as employed?
- At what date was it measured?
- Who is in the denominator?
- Who supplied the data?
Who counted as employed?
At what date?
Who is in the denominator?
Who supplied the data?
| The reported figure | What is left unstated | What to ask |
|---|---|---|
| Employment rate | Definition of employed, and the measurement date | What counted, and when was it counted? |
| Any percentage | Who was excluded from the denominator | What is the denominator, and who is not in it? |
| Salary figure | Base only, or including bonus and other pay | What is included in this figure? |
| An average or median | Which subgroup, and how wide the spread | What does this look like for my function and region? |
| A survey result | The response rate | What share of the class responded? |
| A placement claim | Whether the role required the degree | How many of those roles required this credential? |
A program that publishes definitions, dates, denominators, and response rates alongside its figures is behaving well, and you should say so and take the numbers more seriously. A program that publishes a large percentage with no accompanying definition has told you something, but not about employment.
Salary figures deserve extra suspicion
They are the most quoted and the least comparable. A reported figure may be base salary only or fold in a signing bonus and other compensation. It may cover only the graduates who both found work and chose to report what they earn.
It may be an average pulled upward by a few unusual outcomes, or a median that hides an enormous spread. It almost never separates function, industry, region, or prior experience, and those four variables explain more of the variation in pay than the degree does.
The question that dissolves most salary marketing is simple. For people who did what I intend to do, in the place I intend to do it, with the experience I already have, what does the distribution look like? How many people is that based on? The answer is usually that the program does not know.
For the destination rather than the program, the Bureau of Labor Statistics Occupational Outlook Handbook publishes median pay by occupation. Compare a program's salary claim against that occupation figure before you compare it against anything else. That figure is public, and it is the one to measure a salary claim against.
Official data has its own limits
Government and public data sources are generally more careful than institutional marketing, and they are worth using, but they carry constraints you have to respect.
- They lag.The cohort measured finished years ago, in a different labor market.
- They cover some students and not others, depending on how the dataset was built.
- Field-level figures average across very different programsat very different institutions.
- They describe cohorts, not individuals.The spread within a field is wide, and you are one point in it.
The main federal source is the College Scorecard data site, where institution-level measures are published alongside the documentation that defines them, and what is collected and how is described in the IPEDS collection overview. Read the documentation before the numbers: the definitions decide what any figure means, and they are fixed centrally rather than by each institution's marketing office.
Two fields in that data matter most for this decision. Median earnings shows what former students earn after leaving, and the field of study tables break the same question down by subject. Both are defined in the documentation, and neither matches your personal case.
Use them for orientation and for sanity-checking a claim that seems too good. Do not use them as a personal projection either.
Compare against your own alternative instead
The instinct is to compare a program's numbers against a national average, or against another program's numbers. Both comparisons are close to useless: the first because you are not the average, the second because the two programs almost certainly used different definitions.
The comparison that means something is against your own alternative: what happens if you do not do this.
That alternative is concrete and you can investigate it. It might be staying in your current job for the same number of years and being promoted or not. It might be a cheaper credential, a certification, a shorter course, or a lateral move.
The graduate case, where the alternative is often a stretch assignment rather than another program, is argued in MBA programs. The undergraduate case, where the alternative is usually a different major rather than no degree, is in Business majors.
Compare like this:
Compare against your own alternative
If you do not enrol
- Tuition and fees
- Nothing
- Paid hours given up
- Full pay
- Salary while studying
- $70,000
- Target role
- $85,000 if promoted, chance about one in three
- What you are betting on
- A record you can already see
If you enrol
- Tuition and fees
- $40,000, borrowed
- Paid hours given up
- About five hours a week, worth $10,000 over two years
- Salary while studying
- $70,000
- Target role
- $85,000 if hired, chance about two in three
- What you are betting on
- A credential effect you cannot measure
Program vs your own alternative
This program
- Total cost
- Tuition plus lost earnings
- Three to five years
- Realistic position if you do it
- What must be true
- Assumption that makes it win
- Downside
- If the assumption is wrong
Your alternative
- Total cost
- Cost of the alternative
- Three to five years
- Realistic position if you do not
- What must be true
- Confidence in that assumption
- Downside
- What you fall back on
The arithmetic with those numbers is quick. Tuition of $40,000 plus $10,000 of overtime given up makes the certain cost $50,000 before interest. The salary difference you are buying is $15,000 a year, or $75,000 over five years.
Now weight it by the odds. Moving from one in three to two in three adds one third of $75,000, which is $25,000 in expected salary. That is less than the $50,000 you have already spent.
For the program to break even on those numbers, it would need to move your odds from one in three to close to one, or the pay gap would need to be much wider than $15,000.
Change the inputs to your own situation and the answer may flip. The certain cost is knowable today. The odds usually are not.
If you would have been promoted anyway in those two years, what does the program add?
That last question is the one to spend real time on, because the debt is certain and the outcome is not. The arithmetic of the certain half is in Tuition and financial aid.
How to interrogate a program's own report
Ask in writing, and keep the answers:
- What definition of employment do you use, and on what date do you measure it?
- What is the denominator, and which graduates are excluded from it?
- What was the survey response rate for the most recent cohort?
- Is the data self-reported or verified, and verified how?
- Can you break the figures down by function and region rather than for the class as a whole?
- What share of graduates were sponsored by an employer they returned to?
- How many graduates from the last cohort are working in the specific role I am aiming at?
- Are these figures audited or reviewed by anyone outside the institution?
A short email does the job. Fill the brackets, send it, and keep the reply.
Subject: outcome data request, [program name], [cohort year]
Hello,
I am applying to [program] and I want to read your placement figures correctly. Please confirm in writing:
- the definition of employed you use, and the date you measure it;
- the size of the denominator, and which graduates are excluded from it;
- the survey response rate for the most recent cohort;
- whether the data is self reported or verified, and by whom;
- median earnings by function and region, not just for the class as a whole;
- how many graduates from the last cohort work in the role I am aiming at.
If a number cannot be given, please say so. I will treat that as an answer too.
Thank you.
Notice which questions get a straight answer. That pattern is itself data, and it is more reliable than the numbers.
What an honest program sounds like
It states its definitions before its figures and publishes response rates. It names subgroups it cannot report on because the numbers are too small.
It separates what it provides, teaching, access to a recruiting channel, a credential, from what it cannot provide: a result. Asked what happens to people it does not work for, it answers instead of staying silent.
That candor is a better signal than any percentage, and it is available to you in a single email before you have paid anyone anything. It is also the same test applied to accreditation claims in Business schools: what an institution will put in writing tells you more than what it puts on a page.
